vulkan-samples-staticdev isn't reproducible: https://autobuilder.yocto.io/pub/repro-fail/oe-reproducible-20210109-uud35nir/packages/
Patch in testing and on mailing list
Richard's patch failed on the YP auto-builder. It seems that the debug data in the vulkan binary differs and depends on the path (Debug LOC section ?)
I looked into this issue and whilst we fixed the first problem, there was in fact a second issue where some builds has differing libMachineIndependent.a files. This in turn was caused by differing ParseHelper.cpp.o files. The difference was only in the .debug_loc section, there were no other differences in the binary. In general it was only a couple of loc entries that were different. I also confirmed that an annotated disassemble of ParseHelper.cpp.o was the same for both binaries. After much debugging, I found that on a build which broke on the autobuilder, rebuilding the cmake_pch.hxx.gch file (the precompiled headers) then rebuilding ParseHelper.cpp.o resulted in the correct identical binary being generated. The discrepancy therefore seems to be in the precompiled headers. I tried building the PCH at the start of a build and at the end of the build and also reproducing the exact order of build commands from log.do_compile from the broken build and couldn't reproduce the broken PCH file. In the end I've turned of PCH in cmake for the recipe.
Work around for original issue: http://git.yoctoproject.org/cgit.cgi/poky/commit/?id=49c12484ebb80d72cc84aaba8dc3fc9f2954b4db Also disable PCH: http://git.yoctoproject.org/cgit.cgi/poky/commit/?id=dd1a7c3cabf4b25fbfbd06c91761b88c47b079cc
Fixed on master